When you post to a large public forum like HN, you're broadcasting to thousands. Public broadcasting is not like communicating face-to-face with one person. It's volatile, because the odds of somebody hearing something different from what you had in mind go way up. If you were speaking on, say, television, this would be obvious, but the medium here is more intimate. It feels like personal communication but actually is public communication (or rather, is both), which brings risks. The risks get worse as the community gets larger.
What do to instead? You need to think probabilistically. On the one hand, you're talking to a single person, the person you're replying to. But you're also talking to a statistical cloud of thousands of readers. With the statistical cloud, the question is: what is the probability curve of the reactions the community may have to this comment? If somebody somewhere gets activated, they will rush to the thread to make objections; that in turn will activate others, and off we go. That's how threads go haywire.
You're not responsible for any individual reaction—someone might just be in a bad mood, or have some weird issue—but you are responsible for the "probability curve" of your comment. The expected value of your comment is something like the sum of the areas under that curve: https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...
There's no need to get the probability of a flare-up down to zero. That's impossible. You just need that part of the curve not to be too high. If your comment gets a hostile reaction from one random reader who goes on tilt for whatever reason, that's no big deal because the community won't support it. They'll contain the damage for you. But if the curve is too high, chances are that the community will take it badly—or rather, that some subset will. If the entire community takes it badly, that's also no big deal; the comment will simply be flag-killed. Such posts have a negative value, but it's small. But when one segment of the community takes a comment badly and gets activated, and then a second segment gets activated by their activation, we end up with flamewars—entire subthreads with negative value.
The issue isn't whether you intended such reactions, it's whether your comment would likely give rise to them. That probability is higher if...actually, let me put this positively: that probability is lower if your comment includes markers to disambiguate intent. There are many ways to do that, but the shorter ways are best (e.g. "off topic, but"; or "in case it's helpful"). Even just a little calibration acts as flame retardant. https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...